As enterprises widely adopt generative AI, model invocation volumes are rising rapidly, making token costs a new financial burden. HP believes that one way to reduce these costs is not to continue offloading more tasks to the cloud, but to run part of the AI capability directly on endpoint devices.
AI costs are shifting to the edge
HP’s Chief Strategy and Transformation Officer, Prakash Arunkundrum, told Fortune that every internal conversation, search, and automated invocation consumes tokens, and costs quickly accumulate as usage scales. HP’s assessment is that edge AI can reduce reliance on cloud-based models, thereby lowering the total cost of ownership.
He noted that scenarios such as manufacturing quality inspection are particularly well-suited for local execution. These tasks typically require lower latency and place greater emphasis on keeping data on-site rather than transmitting it to external systems for processing.
AI PCs now account for 35% of sales.
Along this direction, HP is advancing AI PCs, Copilot-enabled printing devices, and smart conference room terminals with intelligent sensing capabilities. The company has also launched its locally prioritized AI model, HP IQ, aiming to connect devices, software, and office scenarios into a unified experience.
HP's latest earnings report shows that revenue for the first quarter of fiscal year 2026 reached $14.4 billion, representing nearly a 7% year-over-year increase. Personal Systems revenue amounted to $10.3 billion, up 11% year over year. Interim CEO Bruce Broussard attributed part of the growth to the continued momentum of AI PCs, noting that these products now account for 35% of device sales.
- First-quarter revenue: $14.4 billion
- Personal system business revenue: $10.3 billion
- AI PCs account for 35% of device sales.
Rising memory prices compress profit margins.
In addition to cost, HP links edge AI with data sovereignty. Arunkundrum noted that in Singapore and across the broader Asian market, some governments place greater emphasis on keeping data within national borders, which can restrict certain cloud-based AI applications. If models can run directly on-device—whether open-source or proprietary—it becomes easier to comply with local regulations and privacy requirements.
However, as HP advances its AI devices, it faces cost pressures due to tight supply of memory chips. The company’s operating profit margin for the first quarter was 5.3%, down from 6.3% in the same period last year. HP previously indicated that profits for this quarter could fall at the lower end of its guidance range, partly due to rising memory costs. The company will announce its second-quarter results on May 27.
Additional information: HP reports that 78% of its employees in the Asia-Pacific region are already using AI in their daily work, higher than its global average of 72%; in October last year, the company also launched Garage 2.0 in Singapore, opening access to engineers, infrastructure, and customer resources for startups.